1 A classification of endangered high-THC cannabis (Cannabis sativa subsp. indica) domesticates and their wild relatives John M. McPartland and Ernest Small Supplementary File Introduction section SF.1. Species concepts p. 1 SF.2. Notes on the other subspecies, C. sativa subsp. sativa p. 6 SF.3. Level nominalism and wild-type nominalism p. 8 Methods section SF.4. Methodology, herbarium studies p. 12 SF.5. Methodology, literature review p. 14 Results section SF.6. Protologues of the four varieties p. 18 SF.7. Nomenclatural priority: debates over Persoon and Cazzuola p. 23 SF.8. Morphological comparisons p. 24 SF.9. Phytochemical comparisons p. 41 SF.10. Molecular genetic comparisons p. 58 Discussion section SF.11. Crossbreeding indica and afghanica into “Sativa” and “Indica” p. 67 SF.12. Intermediate forms; East Asian hemp p. 71 SF.13. Practical applications and future directions p. 74 Representative herbarium specimens p. 76 Reference list for Supplementary File p. 82 Taxonomic abbreviations used in Supplementary File Cannabis = all taxa included in the genus Cannabis C. sativa = Cannabis sativa subsp. sativa var. sativa C. ruderalis = Cannabis sativa subsp. sativa var. spontanea C. indica = Cannabis sativa subsp. indica var. indica C. afghanica = Cannabis sativa subsp. indica var. afghanica C. himalayensis = Cannabis sativa subsp. indica var. himalayensis C. asperrima = Cannabis sativa subsp. indica var. asperrima SF.1. Species concepts At last count, a score of “species concepts” competed for adherents. Four widely accepted species concepts, as they pertain to Cannabis, are elaborated below: The biological species concept (BSC) defines a species as a group of interbreeding populations, reproductively isolated from other groups (Mayr 1942). Mayr tested the “species status” of two organisms with a breeding experiment. If they produce fertile offspring, they are the same species. Small (1972) crossbred 38 Cannabis accessions in a glasshouse experiment. He included fiber-type plants from Europe, Turkey, China, and Japan; drug-type plants from Mexico, Thailand, Syria, Cyprus, and Europe; and wild-type plants from Germany, Canada, and USA (no C. afghanica accessions). All F1 hybrids were interfertile. Small concluded that no sterility barriers existed within the genus, which consisted of one biological species. 2 Some of Small’s hand-pollinated crosses might not naturally hybridize in the field, due to reproductive isolation barriers. Reproductive barriers are expressed along a continuum, as populations diverge from ecotypes, to species, to distinct phylogenetic lineages (Lowry and Gould 2016). The continuum begins with “prezygotic” barriers between populations, which are based on extrinsic mechanisms, and depend on the external environment. At least two prezygotic barriers operate in Cannabis: Temporal (allochronic) isolation arises in the form of separate flowering times. Janischevsky (1924) reported temporal isolation between C. ruderalis and neighboring C. sativa. Wild-type plants matured in mid-June, while domesticated plants were still in the vegetative stage. Temporal isolation thwarted Bredemann (1952) when he tried crossing German C. sativa with South Asian C. indica. By the time Indian males produced pollen, German females had already passed their fertility period. When he pollinated Indian females with German pollen, frost killed Indian females in October before setting seed. Habitat isolation arises in the form of genetic fitness for a specific environment. Habitat isolation can be tested in a transplantation experiment. C. himalayensis adapted to the Himalaya (high altitude, cooler) may not survive when transplanted to the plains of India (low-altitude, hotter). C. indica adapted to the warm-and-wet plains of India would likely succumb in the Hindu Kush, with its arid climate, desiccating winds, high UV-B radiation, and shorter growing period. C. afghanica is poorly adapted to warm-and-wet conditions—the seedlings are susceptible to lethal diseases caused by Pythium and Rhizoctonia fungi in damp soil. In mature plants, roots suffer waterlogging stress, branches snap under monsoonal rainfall, and flowers perish from “bud rot” caused by Botrytis cinerea and Trichothecium roseum. Backcross experiments show that intolerance to humidity is expressed in hybrids that contain a small percentage of C. afghanica parentage (McPartland et al. 2000). The diagnostic species concept (DSC) defines a species as “the smallest group that is consistently and persistently distinct, and distinguishable by ordinary means” (Cronquist 1978). Cronquist applied this concept to Cannabis taxonomy (Small and Cronquist 1976). Folk taxonomists employ a DSC concept when they distinguish between “Sativa” and “Indica”. Distinguishable features are taxonomic characters—attributes of an organism that are divisible into at least two conditions (or states). For example, plant height is a character that distinguishes “Sativa” from “Indica”, with two character states: ≥ 2 m for “Sativa”, and < 2 m for “Indica”. DSC criteria become unreliable with sibling species (which look alike and may be “lumped”), or highly distinctive varieties (which may be “split” into separate species). Taken to the extreme, botanists in the 16th-17th centuries split male and female plants into separate species, calling them Cannabis mas and Cannabis foemina (e.g., Boch 1546). This splitting was based on a single taxonomic character—gender, with two character states—female or male flowers. Modern DSC-based taxonomists measure as many characters as possible, in as many organisms as possible, and give the characters equal weight. This method is known as “numerical taxonomy” or “phenetic” (as opposed to phylogenetic) taxonomy. Phenetic taxonomists apply multivariate statistics, such as principal component analysis (PCA) or neighbor-joining (NJ) 3 methods to discern clusters of organisms, which can be delimited as species or infraspecific groups. Another DSC method, canonical analysis (CA), differs from PCA by pre-defining potential groupings on some criterion. PCA is used for pattern recognition, whereas CA is used for hypothesis testing. CA establishes whether or not a minimal discontinuity of variation exists between groups. Small et al. (1976) and Hillig (2005b) pre-defined their accessions as members of C. sativa, C. indica, or C. ruderalis. Small’s CA analysis did not separate C. ruderalis as a discrete group, whereas Hillig’s CA analysis showed separation. Many Cannabis studies have used NJ, PCA, and CA, which we detail throughout this Supporting Information. See Fig. S1 for a phenetic NJ tree based on the results of this study. Figure S1. Phenetic NJ tree of C. himalayensis Cannabis, with putative C. indica hybridization events marked “Sativa” by dashed lines “Indica” C. afganica C. asperrima “Ruderalis” C. sativa C. ruderalis Humulus lupulus The phylogenetic species concept (PSC) defines a species as an evolutionarily divergent lineage—the smallest set of organisms that share a character state inherited from a common ancestor. Hennig (1966) differentiated between ancestral character states (plesiomorphies), and derived character states (apomorphies). Organisms with deep ancestral roots share plesiomorphies, whereas more recently derived organisms share apomorphies. A shared apomorphy (synapomorphy) is a derived and unique character state, present in two groups of organisms and their last common ancestor, and is not present in earlier ancestors. Prior to Darwin, botanists intuited “primitive” and “advanced” character states, which often correlated with plesiomorphies and apomorphies. Lamarck’s protégé, Augustin de Candolle, first proposed that primitive and advanced character states could be used to organize plant taxonomy. De Candolle (1813) coined the word taxonomie, defined as la théorie des classifications. Bessey (1915) erected the first avowedly evolutionary system to analyze primitive and advanced character states in plants. He composed a list of evolutionary “trends” that indicate the directionality of character changes. He used “trends” to polarize character states—determine which character states were ancestral, and which were derived. Hennig (1966) polarized character states using “outgroup analysis.” He compared a group under study (the ingroup) to its outgroup. For example, when we study Cannabis and Humulus as the ingroup, Celtis can serve as the outgroup. If a character state occurs in the ingroup and the 4 outgroup, it is plesiomorphic—ancestral, shared by distant ancestors. If a character is absent in the outgroup and unique to the ingroup, it is apomorphic. After polarizing a series of characters, Hennig constructed a cladogram—a dichotomously branching diagram that represents a nested hierarchy of ingroups and outgroups. A unique synapomorphy arises at each node, a character state shared by a clade: the common ancestor and its decendants. A simplified example of this process is presented in Fig. S2. Figure S2. Cladogram of four C Cannabis genera in Rosales. Synapomorphies that define B each monphyletic group are Humulus indicated by circled letters, A described in the text. Celtis Morus Fig. S2 illustrates three synapomorphies with respect to the direction of leaf evolution: Synapomorphy A: Leaves with free stipules (monophyletic group: Cannabis, Humulus, Celtis). The ancestral character state is a stipule that ensheaths the petiole (Morus). The stipule is an appendage at the base of the leaf petiole. Synapomorphy B: Leaves with a craspedodromus venation